httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rob Hartill <r...@imdb.com>
Subject Re: idea - last-modified despite SSI - possible performance booster
Date Sun, 02 Feb 1997 02:19:17 GMT
On Sat, 1 Feb 1997, Nathan Neulinger wrote:

> At 12:41 AM +0000 2/2/97, Rob Hartill wrote:
> >On Sat, 1 Feb 1997, Nathan Neulinger wrote:
> >
> >> By returning a timestamp with more of these documents, this could lead to a
> >> major performance increase since the pages could be cached, whereas now
> >> they can't be.
> >
> >You can use "Expires" and "Cache-Control" to make SSI document cacheable.
> >Almost 100% of my HTML output is SSI parsed (again for standard includes
> >but also to insert banner ads).
> >
> >For the standard includes (page footers), I added a PerlHandler
> >(mod_perl) to cache the fixed messages at server startup then intercept
> >the SSI requests for them so they get served from memory not disk. Dunno
> >what it saves, but it gives warm fuzzies just thinking about it.
> 
> Wouldn't that only be usable from the main config files and by the web
> admin? 95% of our web data consists of user pages (where user is a
> student/employee/department/organization)

The idea could be made more general, but that's not what I need so
won't be something I'd do. I could get better efficiency by pulling in
the standard includes on demand before caching them.

Clearly, you wouldn't want a server shared by lots of people being
abused by idiots loading in huge include files for caching in order to
knock the server over.

> For the buffering problem, I was thinking more of whether buffering the
> file and caching would use less CPU than the page being retrieved multiple
> times. My other concern was that it be done transparently. Expires and
> Cache-Control require active action on the users part.

It can be added from the conf files and/or .htaccess or even mod_cern (?)

The advantage of using caching headers is that individual browsers are
saved re-requesting pages they fetched a moment or two ago. I suspect
lots of people have mozilla configured to verify documents once per
session, so people often end up returning to documents within a matter
of seconds/minutes when the server owner would be happy to let them cache it
for N minutes/hours/days/...   I use expires/cache-control to give people
pages for 20 minutes (usually enough time to do cached BACK operations)
before they're "forced" to refetch in order to get a "fresh" ad banner.
(banners are a pain, but they pay the bills and the users get what they pay
for).



rob

Mime
View raw message